BUILD ASRock vs Supermicro + Xeon or FreeNAS Mini?

[STREBLO]

I'm building a server I'm just going to be using at home mainly for backups and media. I might be hosting an owncloud server and using plex as well and I want to have the ability to use it for other things I might not initially plan for so I don't want to under build it.

I'm trying to decide between the ASRock 2750D4I motherboard with SOC octa core atom processor ($485 CAD), and the Supermicro X10SL7-F ($280 CAD) with a Xeon E3-1230v3 ($339 CAD).

Is there any benefit of going with the Supermicro? The ASRock has an octa core processor, although they are Atoms, and is cheaper. Do I gain anything by going with the supermicro.

 

[Ericloewe]

ASRock C2750D4I:

• Lower peak power consumption
• Less powerful
• Only 6 useful SATA ports (and 6 more useless ones)
• Very limited exapndability

Supermicro X10SL7-F + Xeon E3-1230v3:

• Higher peak power consumption
• More powerful
• 14 total immediately usable SATA/SAS ports.
• Better expandability for additional LSI SAS controllers or 10GbE controllers.

 

[jgreco]

On the other hand, the C2750 will still allow you to add an LSI HBA to get support for 14 drives, and that could well be sufficient expandability for home use.

 

[STREBLO]

ASRock C2750D4I:

• Lower peak power consumption
• Less powerful
• Only 6 useful SATA ports (and 6 more useless ones)
• Very limited exapndability

Supermicro X10SL7-F + Xeon E3-1230v3:

• Higher peak power consumption
• More powerful
• 14 total immediately usable SATA/SAS ports.
• Better expandability for additional LSI SAS controllers or 10GbE controllers.

So the Marvell ports are completely useless? How come?

 

[Ericloewe]

So the Marvell ports are completely useless? How come?

Slow, unreliable, their presence adds 10W to board power consumption due to the need for a PCI-e switch...

Take your pick. Some people have had luck with them, but it's a real crapshoot.

 

[STREBLO]

Slow, unreliable, their presence adds 10W to board power consumption due to the need for a PCI-e switch...

Take your pick. Some people have had luck with them, but it's a real crapshoot.

Also, at idle they should have comparable power consumption right? The higher usage will only be when i'm actively using it, am I correct?

 

[jgreco]

So the Marvell ports are completely useless? How come?

Basically because of driver issues. My recollection is that Marvell doesn't provide documentation for their chipset, so that often means that the driver author is left to figure it out.

 

[STREBLO]

Ok, thanks. Guess i'll play it safe with the Supermicro.

 

[jgreco]

Ok, thanks. Guess i'll play it safe with the Supermicro.

If nothing else, the upside to that is that nobody ever comes back and complains that their system is underpowered...

 

[STREBLO]

If nothing else, the upside to that is that nobody ever comes back and complains that their system is underpowered...

By the way, the Xeon E3-1230v3 is probably the best choice, right?

 

[jgreco]

The E3-1231v3 may be slightly better. However, there is not a whole heck of a big difference between the E3-1230, E3-1230v2 (Supermicro X9) and the current 1230v3/1231v3 that you would notice.

 

[STREBLO]

The E3-1231v3 may be slightly better. However, there is not a whole heck of a big difference between the E3-1230, E3-1230v2 (Supermicro X9) and the current 1230v3/1231v3 that you would notice.

Ahh I missed that, guess i'll get the newer one.

 

[STREBLO]

I'm looking at compatable memory and they list DDR3 1600, DDR3 1.35v-1333 and DDR3 1.35v-1600 memory, is it best to get DDR3 1.35v-1600?

 

[INCSlayer]

yea but you probably wont notice the diffrence :P

 

[STREBLO]

yea but you probably wont notice the diffrence :p

But I will KNOW there's a difference ;)

 

[STREBLO]

What exactly is the difference between DDR3 1600 and DDR3 1.35v-1600 memory? Do they just mean any voltage or what by saying DDR3 1600?

 

[INCSlayer]

it has todo with how fast they are 1333 are slower than 1600
dont ask me exactly how ram memory speeds work because i dont really understand it myself i only really know that a higher number is faster and faster = better most of the time

 

[Ericloewe]

Also, at idle they should have comparable power consumption right? The higher usage will only be when i'm actively using it, am I correct?

No, it's constant. That said, it's still comparable to the X10SL7-F.

 

[jgreco]

I'm looking at compatable memory and they list DDR3 1600, DDR3 1.35v-1333 and DDR3 1.35v-1600 memory, is it best to get DDR3 1.35v-1600?

Yes, but not at any price. Your RAM is tremendously faster than what you can cram in and out of a 1Gbps ethernet, or even a 10Gbps ethernet. It would not, for example, be a good idea to pay a premium for 16GB of DDR3 1.35v-1600 if you could get 32GB of DDR3-1333 for a similar price; ZFS makes very good use of RAM, and bumping up the size of the RAM is a LOT more valuable than the faster memory. You might only notice the faster memory if you ran side-by-side benchmarks.

 

[jgreco]

No, it's constant. That said, it's still comparable to the X10SL7-F.

I'm wondering if he meant the *boards*; the X10SL7-F will definitely have higher usage when you're actively using it, but only because it can do a fair bit more. Given a C2750D4I and the X10 board, the X10's roughly twice as fast, so given a set amount of work to do, the X10 will complete it faster and will go back to being idle. Making comparisons between the boards ends up being kind of difficult. Bearing in mind that watts at idle and watts burned for work are not exactly the same for these two boards, if you put the same sort of NAS workloads on these two boxes, a workload that nearly maxes out the C2750 will see the E3 working lazily at it. So the C2750 will be close to its max wattburn, but the E3 won't be anywhere near. If you look at it that way, then they probably consume a vaguely similar number of watts. This may be the measurement that's most meaningful for the average NAS.